Power distribution network topology checking and updating method and system based on visual perception and data fusion

Abstract

This invention discloses a method for verifying and updating the topology of a power distribution network based on visual perception and data fusion, belonging to the field of power distribution network topology verification technology. The method includes: adjusting the drone's pose in real time based on target perception feedback during inspection to actively align with the target; collecting image data of power components containing multimodal information; filtering valid images from the image data and uploading them to a cloud-based management platform; identifying power component entities in the filtered image data and constructing a spatial semantic graph based on the spatial distribution association between components; performing topological logic reasoning and feature consistency analysis on the spatial semantic graph to identify abnormal states of components and determine the power logic attributes of faulty components; and finally generating operation and maintenance instructions based on the identified abnormal states and power logic attributes and pushing them to the terminal. This invention addresses the problems of high data redundancy, ambiguity in the logical identity identification of power components, and difficulty in identifying atypical faults during power distribution network inspection.

Description

Technical Field

[0001] This invention relates to the field of distribution network topology verification technology, and more specifically, to a method and system for verification and updating distribution network topology based on visual perception and data fusion. Background Technology

[0002] The maintenance and verification of power distribution network topology mainly rely on two methods: manual inspection and data entry, and analysis and identification based on electrical measurement data. The manual method typically involves maintenance personnel manually drawing or modifying drawings in the GIS system after project completion or inspection. The method based on electrical measurement data (such as using smart meter AMI data and micro-synchronous phasor measurement unit (uPMU) data) analyzes the time-series correlation or causal relationship of node voltage and current, using graph neural networks or statistical algorithms to deduce electrical connection relationships. In recent years, with the popularization of drone technology, using drone imagery for equipment appearance defect detection has gradually become commonplace, and some advanced solutions have begun to explore the use of image recognition technology to assist in asset inventory.

[0003] However, existing power distribution network topology verification technologies still have significant limitations and blind spots in practical applications. First, manual maintenance methods suffer from severe lag, often resulting in a situation where "the on-site project is already operational, but the system drawings have not been updated," leading to discrepancies between the drawings and reality. Second, mainstream identification methods based on electrical measurement data heavily rely on line load current. For "cold standby" lines or unloaded lines that are disconnected, the lack of electrical signal characteristics often leads algorithms to misjudge them as disconnected, causing physical assets to "disappear" in the digital model and severely impacting the development of emergency power transfer plans. Furthermore, simple electrical simulations cannot perceive actual physical spatial information and are unable to detect physical anomalies such as unauthorized connections or pole / tower position deviations. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for verifying and updating the distribution network topology based on visual perception and data fusion. By constructing a visual perception topology that reflects the physical connections on site, and performing multi-dimensional spatial alignment and logical consistency comparison with the operating and design topologies, the method achieves adaptive updating of the model based on the evaluation results. This solves the problems of delayed model updates and discrepancies between the diagram and reality caused by a single data source or manual maintenance in distribution network management.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The method for verifying and updating the distribution network topology based on visual perception and data fusion includes the following steps: acquiring the operating topology, design topology, and field perception data of the distribution network; extracting entity features and analyzing spatial relationships from the field perception data to construct a visual perception topology; aligning the visual perception topology with the operating topology and design topology using spatial features and comparing their logical consistency to calculate the topology consistency evaluation result; determining the current topology state category of the distribution network based on the topology consistency evaluation result, and performing an adaptive update of the distribution network model using the verified topology data when the preset update trigger conditions are met.

[0007] In a preferred embodiment, constructing the visual perception topology includes: parsing on-site perception data to acquire image data and the pose information and imaging parameters of the acquisition device; identifying power equipment entities in the image data, wherein the power equipment entities include at least towers serving as topology nodes; mapping the image coordinates of the power equipment entities to three-dimensional geographic space coordinates based on the pose information and imaging parameters; analyzing the continuity of conductors crossing adjacent towers in three-dimensional space based on multi-view geometry principles, and determining the conductor connection relationship between adjacent towers by combining the positions of conductor attachment components identified in the image data; and generating the visual perception topology based on the three-dimensional positions of the towers and the conductor connection relationship.

[0008] In a preferred embodiment, the spatial feature alignment and logical consistency comparison includes the following steps:

[0009] Extract the operational and design topologies of the area to be verified and establish a geospatial coordinate system. Project the visually perceived topology and design topology onto the coordinate system of the operational topology. Perform a logical consistency comparison based on the spatially matched data: if the topology consistency evaluation results of the visually perceived topology and the operational topology are consistent, the operational data is determined to be accurate; if they are inconsistent, the design topology is used for verification. If the visually perceived topology and the design topology are consistent, the visually perceived topology is marked as data to be updated; if they are inconsistent, the area is marked as a manually verified object.

[0010] In a preferred embodiment, the topology consistency assessment result is obtained by calculating a comprehensive difference degree, which includes calculating node position deviation, connection similarity, and attribute feature matching degree. The node position deviation degree is the difference between the Euclidean distance or projected distance between the visually perceived node and the system-recorded node in geographic space. The connection similarity is the degree of overlap or set similarity coefficient between the edge set of the visually perceived topology and the edge set of the system topology to be checked. The attribute feature matching degree is the degree of consistency between the identified inherent attributes of the equipment and the attributes recorded by the system. The inherent attributes include tower type, number of insulator strings, or equipment appearance characteristics.

[0011] In a preferred embodiment, the construction of the visual perception topology also includes the identification of the physical state of the switching equipment: identifying the opening and closing characteristics of the sectionalizing switch or the connecting switch in the image data, the characteristics including the position of the operating handle, the direction of the opening and closing indicator, or the contact state; based on the opening and closing characteristics, obtaining the on / off attribute state of the corresponding connecting edge in the visual perception topology; during verification, if the connection relationship between the visual perception topology and the operating topology is consistent but the on / off attribute state is inconsistent, it is determined that the switch state has changed, and a switch state update instruction is generated.

[0012] In a preferred embodiment, the analysis of the continuity of the conductor spanning adjacent towers in three-dimensional space based on multi-view geometry principles further includes verifying the conductor morphology using physical model constraints: constructing a curve mathematical model that conforms to the physical characteristics of the conductor, the curve mathematical model including a catenary model or a parabolic model; calculating the fitting residual or morphological matching degree between the conductor image feature points and the curve mathematical model; and determining whether to establish the conductor connection relationship based on the fitting residual or morphological matching degree.

[0013] In a preferred embodiment, the spatial feature alignment and logical consistency comparison further includes identifying the status of cold standby lines through electrical measurement data, including: when the electrical measurement data indicates that the line is in a state of no current or zero load, auxiliary judgment is made based on visual perception topology: if the visual perception topology shows that the line has a conductor connection relationship and the switching equipment on the connection path is in a disconnected state, the line is determined to be in a cold standby state, and the connection relationship is retained and the electrical status is marked as disconnected during the update; if the visual perception topology shows that the line does not have a conductor connection relationship, the line is determined to be physically disconnected or removed, and the connection is removed from the running topology data during the update.

[0014] In a preferred embodiment, the design state topology acquisition method is as follows: parsing the electronic design document to extract the geometric coordinate information and attribute text information of the graphic elements; using spatial clustering or graphic element connection analysis algorithms to reorganize the discrete line graphics elements into an engineering drawing structure containing topological connection relationships; establishing a transformation mapping relationship between the design coordinate system and the geographic coordinate system, and transforming the reorganized design state topology to a geographic coordinate system unified with the operational state topology.

[0015] In a preferred embodiment, before performing the adaptive update of the distribution network model using the verified topology data, a multi-dimensional confidence verification step is further included: calculating the comprehensive confidence score of the topology area to be updated, the score being calculated based on the degree of verification consistency, visual image clarity, and feature extraction density; only when the comprehensive confidence score reaches a preset confidence threshold is the automated update operation performed, otherwise an audit report containing on-site images and difference annotations is generated and pushed to a human interaction terminal for confirmation.

[0016] This invention provides a distribution network topology verification and update system based on visual perception and data fusion, comprising: a data acquisition module for acquiring the operating topology, design topology, and field-sensed data of the distribution network; a topology construction module for extracting entity features and analyzing spatial relationships from the field-sensed data to construct a visually perceived topology; a topology analysis module for aligning the spatial features and comparing the logical consistency of the visually perceived topology with the operating topology and design topology, respectively, and calculating a topology consistency evaluation result; and a topology update module for determining the current topology state category of the distribution network based on the topology consistency evaluation result, and, when a preset update trigger condition is met, performing an adaptive update of the distribution network model using the verified topology data.

[0017] The technical effects and advantages of the present invention regarding the method for verifying and updating the distribution network topology based on visual perception and data fusion are as follows:

[0018] This invention achieves automated verification and adaptive updating of distribution network topology by integrating field-sensed data, operational topology data, and design topology data. Specifically, it utilizes visual perception technology to construct a visually perceived topology reflecting the physical connections on-site. This topology is then compared spatially and logically across multiple dimensions with the system-recorded operational topology and the theoretically planned design topology to quantitatively assess topology consistency. Based on this consistency assessment, the system can accurately determine the true state of the current topology and automatically update the system model using the verified accurate topology data when update conditions are met. This invention significantly solves the problems of delayed topology updates and discrepancies between maps and reality caused by single data sources, excessive manual intervention, or insufficient sensing capabilities in traditional methods, thereby improving the automation level and data accuracy of distribution network management. Attached Figure Description

[0019] Figure 1 A schematic diagram of the method for verifying and updating the topology of a power distribution network based on visual perception and data fusion provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of fitting the conductor feature points and the catenary model provided in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the comprehensive confidence score distribution and automated threshold segmentation provided in an embodiment of the present invention;

[0022] Figure 4 This is a block diagram of the distribution network topology verification and update system based on visual perception and data fusion provided in an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1, Figure 1 This invention presents a method for verifying and updating the topology of a power distribution network based on visual perception and data fusion, comprising the following steps:

[0025] S1 acquires the operating topology, design topology, and field sensing data of the distribution network.

[0026] It should be noted that the operational topology of the distribution network is usually extracted directly from the database of the power company's Production Management System (PMS) or Geographic Information System (GIS). The data format is a CIM (Common Information Model) file, SVG graphic exchange format file, or Shapefile geospatial data file that conforms to the IEC 61968 / 61970 standard. It contains the equipment ledger information and connection relationships currently recorded in the system. Field perception data is obtained by drones or ground inspection robots equipped with high-precision RTK positioning modules and visible light cameras. Specifically, it includes high-resolution inspection image sequences covering the area to be verified and the POS information (including longitude, latitude, altitude, roll angle, pitch angle, and yaw angle) corresponding to each frame of the image. Design topology data usually exists in the form of unstructured electronic engineering drawings and needs to undergo specific digital analysis and conversion processing before it can be used for subsequent comparison.

[0027] In this embodiment, the method for obtaining the design-state topology is as follows:

[0028] 1) Parse electronic design documents to extract geometric coordinates and attribute text information of graphic elements. Specifically, use a CAD parsing engine to read DWG or DXF format engineering drawing files provided by the design department and traverse the entity database within the file. Classify and extract graphic elements, identifying specific blocks or closed loops defined in the drawings as "node elements" representing equipment such as towers and transformers, and extracting their insertion point coordinates as local geometric locations. Identify polylines, straight lines, or spline curves in the drawings as "edge elements" representing conductors, and extract their start, end, and inflection point coordinate sequences. Simultaneously, extract text or multi-line text entities from the layers to obtain attribute information such as equipment name, pole number, and line model. Based on the principle of spatial proximity, associate the attribute text with the node element or edge element closest to its geometric center, forming a discrete geometric dataset with attribute labels.

[0029] 2) Using spatial clustering or primitive connection analysis algorithms, discrete line primitives are reorganized into an engineering drawing structure containing topological connections. Due to drawing errors or paper layering, visually connected lines may have minor breaks at the data level. Therefore, this embodiment uses a spatial clustering algorithm based on a distance threshold to construct the topology. The specific steps are as follows: Set a capture threshold (e.g., 5 pixels in paper units). For the endpoint of each edge primitive, search the set of node primitives for candidate nodes whose Euclidean distance is less than the capture threshold. If a candidate node exists, establish a logical connection between the edge and the node. If there are no node primitives near the endpoint of an edge primitive, but there is an endpoint of another edge primitive with a distance less than the capture threshold, it is determined to be a line segment or T-junction, a virtual node is generated, and the two edges are logically merged. Through the above process, unstructured geometry is transformed into a set of nodes. Sum of edges Graph structure data .

[0030] 3) Establish a transformation mapping relationship between the design coordinate system and the geographic coordinate system, transforming the reorganized design topology to a geographic coordinate system unified with the operational topology. Since design drawings typically use an independent local Cartesian coordinate system, while the operational topology uses a national geodetic coordinate system (such as CGCS2000), registration using control points is necessary. Specifically, select at least three distinctive graphic elements (such as substation outlet towers or obvious road intersections) as control points in the design drawings, find the corresponding geographic coordinate points on the GIS map, and use the least squares method to solve for the two-dimensional affine transformation parameters or the four-parameter Helmert transformation model.

[0031] The conversion formula between the design coordinate system and the geographic coordinate system is as follows:

[0032] (1)

[0033] In the formula, The transformed coordinate vectors are unified to the geographic coordinate system of the running topology. The original local coordinate vector extracted from the design drawings; This is a scaling factor used to correct for differences between the drawing scale and the actual geographical scale. This is the rotation angle, used to correct the deviation between the true north direction on the drawing and the geographical true north direction; and These represent the east-west and north-south translations, respectively. This transformation ensures that the design-state topology, operational topology, and field-sensed data are aligned with the same spatial reference for subsequent feature alignment.

[0034] This step uses automated parsing and spatial transformation technology to convert unstructured CAD drawing data into a computer-understandable graph theory topology structure and achieves precise alignment with GIS geographic information. This provides a reliable data foundation for the subsequent introduction of "design state" as a verification benchmark, avoiding the logical blind spots that exist when relying on a single data source for verification.

[0035] S2 extracts entity features and analyzes spatial relationships from the on-site perception data to construct a visual perception topology.

[0036] It should be noted that traditional distribution network topology maintenance mainly relies on manual on-site verification followed by manual data entry into the system, or on logical deduction based solely on the electrical on / off data of smart meters. The former is inefficient and prone to "data entry time lag," while the latter cannot perceive the actual physical connections, such as distinguishing between open circuits and circuit breaker trips, nor can it determine the specific spatial orientation, leading to frequent discrepancies between the map and reality. The solution in this step uses computer vision technology to directly extract the "entity-relationship" structure from physical images, constructing an objective visual topology with geographic coordinate constraints, serving as the "truth value" for verification.

[0037] In this embodiment, a deep learning-based object detection algorithm is first used to extract pixel features of key power entities such as poles and conductors from massive inspection images. These discrete pixels are then mapped to a real-world three-dimensional geographic coordinate system using photogrammetry principles, thus constructing the geometric framework of the power distribution network. Building upon this, a multi-view geometric constraint and physical-mechanical model verification mechanism is introduced to address the continuity determination and anti-interference issues of conductor span connections, ensuring the physical authenticity of the topology connections. Simultaneously, a deep fusion of visual semantic understanding of the switching equipment's on / off states is used to assign dynamic electrical on / off attributes to the topology edges. Through this hierarchical processing from coarse to fine, from geometric to semantic, a high-confidence visual perception topology map is ultimately generated, containing both precise spatial location and real-time operating status, providing a solid factual basis for subsequent heterogeneous data fusion verification.

[0038] Specifically, the construction of the visual perception topology includes:

[0039] 1) Analyze the on-site sensing data to obtain image data and the pose information and imaging parameters of the acquisition equipment.

[0040] 2) Identify power equipment entities in the image data, wherein the power equipment entities include at least poles and towers as topological nodes; specifically, the system employs an improved deep convolutional neural network model, such as YOLOv8 or Mask R-CNN, as a detector to perform inference on the input UAV inspection images. The model is trained to identify multiple types of targets, where "poles and towers" and "transformers" are identified as bounding boxes, serving as node objects in the topology.

[0041] 3) Based on pose information and imaging parameters, the image coordinates of the power equipment entity are mapped to three-dimensional geographic space coordinates. Since a single image can only provide two-dimensional pixel coordinates, this embodiment uses the collinearity equation principle of photogrammetry, combined with the latitude, longitude, elevation and attitude angles (roll, pitch, yaw) recorded by the UAV POS system at the shooting time, as well as the camera intrinsic parameters (focal length, principal point), to establish the mapping relationship between the pixel coordinate system and the geocentric coordinate system (or the local tangent plane coordinate system).

[0042] The formula for calculating the collinearity equation that maps the image coordinates to spatial direction vectors is as follows:

[0043] (2)

[0044] In the formula, The pixel coordinates of the identified power equipment feature points on the image plane; Let these be the coordinates of the principal point; The focal length of the camera; The coordinates of the power equipment to be determined are in the geographic space system. The spatial coordinates of the camera's optical center at the moment of shooting; For the camera attitude angle ( The elements in the rotation matrix formed by the image are used to solve the collinearity equations of multiple images of the same object taken from different viewpoints, and then the bundle adjustment or multi-view adjustment is performed. Figure 3 Angle measurement methods can be used to calculate the precise three-dimensional geographic coordinates of the target object. .

[0045] 4) Based on the principle of multi-view geometry, analyze the continuity of conductors crossing adjacent towers in three-dimensional space, and determine the conductor connection relationship between adjacent towers by combining the positions of conductor attachment parts identified in the image data. Specifically, for situations where a single photo cannot cover the entire view due to large spans of conductors, use epipolar constraints and feature matching algorithms such as SIFT or SuperGlue to track the same conductor in a continuous image sequence. When it is detected that one end of the conductor is connected to the insulator attachment point of tower A, and the extension trajectory of the conductor intersects with the insulator attachment point of tower B in three-dimensional space or the distance is less than a preset threshold (e.g., 0.5 meters), it is determined that there is a conductor connection relationship between tower A and tower B.

[0046] 5) Based on the set of nodes corresponding to the identified 3D positions of the towers and the set of wire connection edges corresponding to the wire connection relationships, a visual perception topology is generated. The final output is graph structure data. ,in Includes the IDs and 3D coordinates of all identified towers. It includes the connection relationships and length information of all wires.

[0047] The construction of the visual perception topology also includes the identification of the physical state of switching devices:

[0048] 1) Identify the opening and closing characteristics of sectionalizing switches or connecting switches in the image data. These characteristics include the position of the operating handle, the direction of the opening / closing indicator, or the contact state. Specifically, when a pole-mounted switch, such as a circuit breaker, load switch, or drop-out fuse, is detected on the tower, a refined identification submodule is triggered. For pole-mounted circuit breakers, an image classification algorithm is used to identify the color of their mechanical pointers or indicator, for example, "red / vertical" represents closed, and "green / horizontal" represents open. For drop-out fuses or disconnectors, the determination is made by detecting the angle or geometric distance between the moving and stationary contacts.

[0049] 2) Based on the opening and closing characteristics, obtain the on / off attribute state of the corresponding connecting edge in the visual perception topology; if the recognition result is "closed", mark the corresponding topology edge as State=1 (conducting); if the recognition result is "open" or "falling", mark it as State=0 (disconnected).

[0050] 3) During verification, if the connection relationship between the visual perception topology and the operational topology is consistent but the on / off attribute status is inconsistent, it is determined that the switch status has changed, and an update instruction is generated to correct the switch status in the operational model. For example, if the system displays that a certain interconnection switch is open, but the visual perception shows that its indicator is "closed", it is determined that the system status has not been updated in time, and a change correction instruction is generated.

[0051] The analysis of the continuity of conductors spanning adjacent towers in three-dimensional space based on multi-view geometry principles also includes verifying the conductor morphology using physical model constraints:

[0052] 1) Construct a curve mathematical model that conforms to the physical characteristics of the conductor. The curve mathematical model includes a catenary model or a parabolic model. In order to eliminate false detections caused by linear interference objects such as tree branches, wall edges, or road markings, this embodiment introduces a mechanical model of a flexible cable as a priori constraint. Since overhead transmission lines exhibit a catenary shape under their own weight, a catenary model is used to represent them.

[0053] The mathematical expression for the catenary model is as follows:

[0054] (3)

[0055] In the formula, The local coordinates of the traverse in the vertical projection plane; The weight per unit length of the conductor; The horizontal tension at the lowest point of the conductor; This represents the horizontal coordinate offset of the lowest point of the conductor. This represents the vertical coordinate offset. When the span is small or the height difference is not significant, this model can also be simplified to a quadratic parabola model for approximate calculation.

[0056] 2) Calculate the fitting residuals or morphological matching degree between the feature points of the traverse image and the mathematical model of the curve; in practice, any one or two of the following indicators can be selected for calculation based on computing resources and accuracy requirements:

[0057] Indicator 1: Fitting Residual. The least squares method is used to project the identified set of conductor sampling points onto the fitted curve, and the root mean square error (RMSE) from the sampling points to the theoretical curve is calculated.

[0058] The fitting residual The calculation formula is as follows:

[0059] (4)

[0060] In the formula, The number of sampling points. For the first The actual ordinate of each sampling point The theoretical ordinate is calculated based on the catenary model. The smaller the value, the smaller the absolute error. For example... Figure 2 As shown in the figure, the fitting effect between the discrete traverse feature points collected in the field and the constructed catenary physical model is illustrated. The scattered points represent the original traverse feature points extracted using a visual algorithm, while the solid line represents the theoretical physical shape obtained by regression based on the catenary formula. It can be seen that the actual traverse point set is closely distributed around the theoretical curve, and the small vertical distance between them represents the fitting residual.

[0061] Indicator 2: Morphological Matching. The similarity between the conductor morphology and the physical model is characterized by the coefficient of determination in statistics or a normalized score based on Fréchet distance.

[0062] In this embodiment, the coefficient of determination is used as the morphological matching degree. The calculation formula is as follows:

[0063] (5)

[0064] In the formula, This is the average of the ordinates of all sampled points. The range of values ​​is The closer the value is to 1, the more the identified line shape matches the physical characteristics of a catenary (i.e., the higher the shape matching degree).

[0065] 3) Determine whether to establish the wire connection relationship based on the fitting residual or morphological matching degree. The system presets a judgment threshold; if the calculated index meets any of the following conditions:

[0066] Condition A: Fitting residuals (For example rice);

[0067] Condition B: Morphological matching degree (For example );

[0068] If the identified object meets the physical characteristics of a flexible conductor, it is confirmed as a real conductor, and a physical topology connection is established between adjacent towers; otherwise, it is determined to be a non-conductor interference object (such as tree branches or straight building edges) and is removed in the topology construction.

[0069] This step, by integrating deep learning object detection, photogrammetric localization, and physical-mechanical model verification, achieves automated conversion from non-contact image data to high-precision 3D topology maps. In particular, the introduced on / off state visual recognition and catenary morphology verification mechanisms not only solve the topology perception blind spot in the "wired but no-electricity" state, but also significantly improve the anti-interference ability of visual recognition in complex backgrounds, ensuring that the generated visually perceived topology has extremely high physical authenticity and credibility.

[0070] S3. The visually perceived topology is aligned with the operational topology and the design topology in terms of spatial features and logical consistency, and the topology consistency evaluation result is calculated.

[0071] In this embodiment, step S3 first unifies the coordinate system to overlay topology layers from different dimensions onto the same geographic space; second, it uses a comprehensive difference algorithm to quantify the deviation between the visual perception results and the system records; finally, it introduces electrical measurement data and design drawings as auxiliary criteria to achieve accurate classification and handling of complex working conditions such as "late update", "illegal construction" and "cold standby".

[0072] Specifically, the spatial feature alignment and logical consistency comparison includes the following steps:

[0073] 1) Extract the operational and design topologies of the area to be verified and establish a geospatial coordinate system. Project the visually perceived topology and design topology onto the coordinate system of the operational topology. Specifically, using the national geodetic coordinate system (such as CGCS2000) or local independent coordinate system adopted by the operational topology as a reference, and utilizing the coordinate transformation parameters obtained in S1 and S2, perform affine transformation projection on the node coordinates of the visually perceived topology and the primitive coordinates of the design topology. During this process, the system uses a kd-tree-based nearest neighbor search algorithm to transform the nodes in the visual topology... Nodes in the running topology Pairing is performed by setting a maximum search radius (e.g., 5 meters). If a corresponding system node exists within this radius, a mapping pair is established. .

[0074] 2) Consistency comparison of execution logic based on spatially matched data:

[0075] If the topology consistency assessment results of the visual perception topology and the operational topology are consistent, the operational data is considered accurate. If they are inconsistent, the design-state topology is used for verification. The specific judgment logic is as follows: the system first calculates the consistency score between the visual perception topology and the operational topology. If the score is higher than the preset confidence threshold (e.g., 0.9), the current system drawings are considered to match the actual scene and no update is required. If the score is lower than the threshold, it indicates that there is a discrepancy between the drawings and the actual scene. At this time, the system automatically calls the design-state topology as the "theoretical benchmark" to intervene in the arbitration.

[0076] If the visually perceived topology matches the design topology, the visually perceived topology is marked as data to be updated; otherwise, the area is marked as a manually checked object. This logic covers three typical scenarios in distribution network operation and maintenance, specifically:

[0077] Scenario A (Normal State): The visual perception topology is consistent with the runtime topology. This indicates that the system data is accurate and no action is required.

[0078] Scenario B (Lag State): The visual perception topology is inconsistent with the runtime topology, but consistent with the design state topology. and This indicates that the site has been constructed according to the design drawings, but the GIS system has not yet completed the as-built data entry process. At this time, the system automatically adopts visual data to correct the model.

[0079] Scenario C (Abnormal State): The visual perception topology is inconsistent with both the runtime topology and the design topology. and This indicates that there may be unreported illegal construction or unfiled design changes on site. The system generates a red alert and pushes it to the manual verification terminal.

[0080] The topology consistency assessment result is obtained by calculating the comprehensive difference degree, which includes calculating the node position deviation degree, connection relationship similarity degree, and attribute feature matching degree. In order to quantify the degree of "consistency" and "inconsistency", this embodiment constructs a multi-dimensional scoring system.

[0081] The node position deviation is the difference between the Euclidean distance or projected distance between the visually perceived node and the system-recorded node in geographic space; this indicator reflects the degree of conformity between the device's geographical location and its location.

[0082] The node position deviation score The calculation formula is as follows:

[0083] (6)

[0084] In the formula, The number of successfully matched node pairs; and The first The three-dimensional coordinate vectors of the matching nodes in the visual perception topology and the system runtime topology; Indicates Euclidean distance; For distance attenuation coefficient (e.g., take...) ), used to map distance deviation to Within an interval, the smaller the deviation, the closer the score is to 1.

[0085] The connection similarity is the degree of overlap or set similarity coefficient between the edge set of the visually perceived topology and the edge set of the topology to be checked in the system; this index reflects the degree of conformity of the network topology, and is specifically calculated using the Jaccard similarity coefficient.

[0086] The similarity score of the connection relationship The calculation formula is as follows:

[0087] (7)

[0088] In the formula, Let be the set of edges in the visually perceived topology. The set of edges in the running topology; Represents the number of elements in a set; intersection operation Union operation represents a common edge in two topologies where both ends of the topology are matched and connected; This represents all the edges involved.

[0089] The attribute feature matching degree is the degree of consistency between the identified inherent attributes of the equipment and the attributes recorded by the system. The inherent attributes include tower type, number of insulator strings, or equipment appearance characteristics.

[0090] The attribute feature matching score The calculation formula is as follows:

[0091] (8)

[0092] In the formula, The total number of attribute items participating in the comparison; For the Kronecker function, when visual recognition attributes With system record attributes If the values ​​match, set the value to 1; otherwise, set the value to 0.

[0093] The final overall difference (i.e., the consistency assessment result) is obtained by combining the above three dimensions. The calculation formula is as follows:

[0094] (9)

[0095] In the formula, Let be the weight coefficients of each sub-indicator, and satisfy . .when If the value exceeds the preset consistency threshold (e.g., 0.85), the two topologies are considered to be consistent.

[0096] The spatial feature alignment and logical consistency comparison also include identifying the status of cold standby lines through electrical measurement data, including:

[0097] When determining that a line is in a state of no current or zero load based on electrical measurement data, visual perception topology is used for auxiliary judgment. It should be clarified that in this embodiment, "no current state" is defined as the measured current of the line. Less than the preset zero drift threshold (For example For this type of line, traditional pure data verification would misjudge it as "no connection," while this solution introduces visual perception for logical verification, the specific logic of which is described below:

[0098] If the visual topology display shows that there are wire connections in the line and the switch on the connection path is in the open state, the line is determined to be in a cold standby state. During the update, the connection is retained and the electrical status is marked as open. This situation corresponds to "there is a physical connection, no electrical current and the switch is open", which confirms that the line exists as a backup power source or is disconnected for maintenance, and is not a system topology error.

[0099] If the visual topology display shows no wire connection in the line, the line is determined to be physically broken or removed, and the connection is removed from the running topology data during the update. This situation corresponds to "no electrical current and no physical connection," confirming that the line entity no longer exists and is redundant data that should be deleted. The specific multi-dimensional state determination logic is shown in Table 1. The table lists the system's determination of the final state of the line under different combinations of electrical measurement states, visual topology connection states, and switch physical states.

[0100]

[0101] This step effectively addresses two major industry pain points in distribution network topology verification: "asynchronous design changes" and "unobservable cold standby assets," by constructing a multi-dimensional quantitative evaluation system encompassing "space, logic, and attributes" and creatively introducing a "visual-electrical" bimodal truth table logic. It not only detects explicit topology errors but also identifies implicit operational inconsistencies, significantly improving the accuracy and self-healing capabilities of the distribution network model.

[0102] S4. Based on the topology consistency assessment results, determine the current topology state category of the distribution network, and when the preset update trigger condition is met, use the verified topology data to perform adaptive update of the distribution network model.

[0103] It should be noted that this step is crucial in converting the physical truth values ​​perceived by the front end into system data. Based on the evaluation results output in S3, the area to be verified is divided into a "consistency zone," a "zone to be updated," and an "abnormal alarm zone." For the "zone to be updated," to avoid frequent, scattered modifications interfering with distribution network scheduling, this embodiment adopts an evolutionary update strategy based on the entire lifecycle of the project. Specifically, the system interfaces with the distribution network project management system to monitor the project progress status in real time. When a project area (such as a supporting project for a new residential area) enters a specific time node of "completion acceptance" or "eve of commissioning," an update condition is triggered. At this time, the system encapsulates the verified visually perceived topology into a standardized CIM / XML diagram update package and uses a database transaction mechanism to atomically replace the old topology in the running database. That is, either all updates succeed, or all are rolled back in case of errors, ensuring data consistency. At the same time, the system automatically retains a snapshot of the historical version before the update so as to support one-click rollback in case of unexpected events.

[0104] In this embodiment, before performing the adaptive update of the distribution network model using the verified topology data, a multi-dimensional confidence verification step is also included:

[0105] A comprehensive confidence score is calculated for the topology region to be updated. This score is based on a weighted average of verification consistency, visual image clarity, and feature extraction density. To prevent data contamination due to blurry images or false alarms caused by severe weather, this embodiment introduces a quantitative "security access control" mechanism before performing the write operation. The system comprehensively considers three dimensions—"result credibility (consistency)," "source data quality (clarity)," and "process completeness (feature density)"—to calculate the final score.

[0106] The overall confidence score The calculation formula is as follows:

[0107] (10)

[0108] In the formula, The topology consistency evaluation result calculated in step S3 (normalized to 0-100 points) reflects the degree of consistency between visual topology and design / physical logic; The visual image sharpness score is obtained by calculating the Laplacian variance or Tenengrad gradient function of the image, and is used to evaluate whether the image has motion blur or focus failure. The feature extraction density score is specifically the ratio (after normalization) of the number of identified effective power primitives to the area per unit geographic area, reflecting the richness of the perceived data. These are the weighting coefficients for the three indicators mentioned above. In the preferred configuration of this embodiment, to highlight the importance of the verification result itself while also considering the quality of the data source, the weight allocation is set as: consistency weight. (50%) Image sharpness weighting (30%), feature extraction density weight (20%).

[0109] The automated update operation is only performed when the overall confidence score reaches a preset confidence threshold; otherwise, an audit report containing on-site images and discrepancy annotations is generated and pushed to a human-interactive terminal for confirmation. The specific implementation logic is as follows: a strict automated entry threshold is set. (For example point).

[0110] when When the system determines that the topology correction data for the region is highly reliable and the source data is of excellent quality, it directly triggers the automated data entry process, updates the running model, and records logs.

[0111] when When the system detects a potential risk (such as slightly blurred images, sparse feature points, or minor logical conflicts), it suspends the automatic update process and generates a visual discrepancy review report. This report displays the "System Original Image," "Visual Reconstruction Image," and "On-site Original Image" side-by-side, highlighting discrepancies with bold borders. This report is then pushed to the desktop or mobile terminals of maintenance personnel via a message queue, awaiting manual "one-click confirmation" or "reject and re-acquire" actions. Figure 3 As shown in the figure, the distribution of the overall confidence score of the topology region to be updated and the boundary of the automated processing decision are illustrated. The horizontal axis represents the overall confidence score (0-100 points), and the vertical axis represents the number of topology units to be processed. The dashed line indicates the preset automated admission threshold. High-confidence samples (green area) located to the right of the dashed line will be automatically taken over by the system and updated into the database, while low-confidence samples (yellow area) located to the left of the dashed line will be intercepted and transferred to the manual review process.

[0112] This embodiment, by constructing a hierarchical processing mechanism of machine initial review and manual verification, not only fully utilizes the efficiency advantages of AI automation, but also safeguards against potential algorithmic risks through a strict confidence threshold, ensuring the absolute security and accuracy of core operating data of the power distribution network.

[0113] Example 2, Figure 4 A distribution network topology verification and update system based on visual perception and data fusion is presented, including:

[0114] The data acquisition module is used to acquire the operating topology, design topology, and field sensing data of the distribution network.

[0115] The topology construction module is used to extract entity features and analyze spatial relationships from field perception data to construct a visual perception topology.

[0116] The topology analysis module is used to perform spatial feature alignment and logical consistency comparison between the visually perceived topology and the running topology and the design topology, respectively, and calculate the topology consistency evaluation result.

[0117] The topology update module is used to determine the current topology status category of the distribution network based on the topology consistency assessment results, and to perform adaptive updates of the distribution network model using the verified topology data when the preset update trigger conditions are met.

[0118] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0119] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0120] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0121] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0122] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0123] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for verifying and updating the topology of a distribution network based on visual perception and data fusion, characterized in that, Includes the following steps: Acquire operational topology, design topology, and field sensing data of the distribution network; Entity feature extraction and spatial relationship analysis are performed on the on-site perception data to construct a visual perception topology; The visual perception topology is aligned with the operational topology and the design topology in terms of spatial features and logical consistency. The topology consistency evaluation result is calculated. Specifically, the operational topology and the design topology of the area to be checked are extracted, and a geospatial coordinate system is established. The visual perception topology and the design topology are projected onto the coordinate system of the operational topology to achieve spatial feature alignment. If the topology consistency assessment results of the visual perception topology and the operational topology are consistent, the operational data is determined to be accurate. If they are inconsistent, the design topology is used for verification. If the visual perception topology is consistent with the design topology, the visual perception topology is marked as data to be updated. If they are inconsistent, an anomaly is determined to exist on site, and an anomaly alarm message is generated. The spatial feature alignment and logical consistency comparison also includes identifying the line status by combining electrical measurement data: when the electrical measurement data indicates that the line is in a no-current or zero-load state, if the visual perception topology shows that there is a wire connection relationship and the switching equipment is identified as being in an open state, then the line is determined to be in a cold standby state, and the connection relationship is retained and the electrical status is marked as open during the update; if the visual perception topology shows that there is no wire connection relationship, then it is determined to be a physical disconnection and the connection is removed from the running topology data during the update. The topology status category of the current distribution network is determined based on the topology consistency assessment results, and the adaptive update of the distribution network model is performed using the verified topology data when the preset update trigger conditions are met.

2. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 1, characterized in that, The construction of the visual perception topology includes: Analyze on-site sensing data to obtain image data and the pose information and imaging parameters of the acquisition equipment; Identify power equipment entities in image data, wherein the power equipment entities include at least poles and towers as topology nodes; Based on pose information and imaging parameters, the image coordinates of the power equipment entity are mapped to three-dimensional geospatial coordinates; Based on the principle of multi-view geometry, the continuity of conductors crossing adjacent towers in three-dimensional space is analyzed, and the connection relationship between conductors between adjacent towers is determined by combining the position of conductor splicing components identified in the image data. A visual perception topology is generated based on the three-dimensional position of the tower and its connection with the conductor.

3. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 1, characterized in that, The topology consistency assessment result is obtained by calculating the comprehensive difference degree, which includes calculating the node position deviation degree, connection relationship similarity degree, and attribute feature matching degree. The node position deviation is the difference between the Euclidean distance or projected distance between the visually perceived node and the system-recorded node in geographic space. The similarity of the connection relationship is the degree of overlap or set similarity coefficient between the edge set of the visual perception topology and the edge set of the topology to be checked in the system. The attribute feature matching degree is the degree of consistency between the identified inherent attributes of the equipment and the attributes recorded by the system. The inherent attributes include tower type, number of insulator strings, or equipment appearance characteristics.

4. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 2, characterized in that, The construction of the visual perception topology also includes the identification of the physical state of switching devices: Identify the opening and closing characteristics of sectionalizing switches or interconnecting switches in image data, including the position of the operating handle, the direction of the opening / closing indicator, or the contact state of the contacts; Based on the opening and closing characteristics, the on / off attribute status of the corresponding connecting edge in the visual perception topology is obtained; During the verification process, if the connection relationship between the visual perception topology and the operational topology is consistent but the on / off attribute states are inconsistent, it is determined that the switch state has changed, and a switch state update instruction is generated.

5. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 2, characterized in that, The analysis of the continuity of conductors spanning adjacent towers in three-dimensional space based on multi-view geometry principles also includes verifying the conductor morphology using physical model constraints: Construct a mathematical model of a curve that conforms to the physical properties of a conductor, wherein the mathematical model of the curve includes a catenary model or a parabola model; Calculate the fitting residuals or morphological matching degree between the feature points of the conductor image and the mathematical model of the curve; Whether to establish the wire connection relationship is determined based on the fitting residual or the morphological matching degree.

6. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 1, characterized in that, The method for obtaining the design state topology is as follows: Analyze electronic design documents to extract geometric coordinate information and attribute text information of graphic elements; Using spatial clustering or primitive connection analysis algorithms, discrete line primitives are reorganized into an engineering drawing structure containing topological connections; Establish a transformation mapping relationship between the design coordinate system and the geographic coordinate system, and transform the reorganized design topology to a geographic coordinate system that is consistent with the operational topology.

7. The method for verifying and updating the distribution network topology based on visual perception and data fusion according to claim 1, characterized in that, Before performing adaptive updates of the distribution network model using the verified topology data, a multi-dimensional confidence verification step is also included: Calculate the overall confidence score of the topology region to be updated, which is obtained by weighting the degree of verification consistency, visual image clarity, and feature extraction density. The automated update operation is only performed when the overall confidence score reaches the preset confidence threshold; otherwise, an audit report containing on-site images and difference annotations is generated and pushed to the human interaction terminal for confirmation.

8. A system using the distribution network topology verification and update method based on visual perception and data fusion as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire the operating topology, design topology, and field sensing data of the distribution network. The topology construction module is used to extract entity features and analyze spatial relationships from field perception data to construct a visual perception topology. The topology analysis module is used to perform spatial feature alignment and logical consistency comparison between the visually perceived topology and the running topology and the design topology, respectively, and calculate the topology consistency evaluation result. The topology update module is used to determine the current topology status category of the distribution network based on the topology consistency assessment results, and to perform adaptive updates of the distribution network model using the verified topology data when the preset update trigger conditions are met.

Images